Episode 103 ... Dave Bergstrom (55:32)

Realized he didn't want to go to law school halfway through undergrad
Father worked for Del Monte foods
Thought he could get rich quick by trading
[Sounds like a southerner or Texan?]
Got into trading around, just after, Great Financial Crisis
Always been a hustler, used to sell counterfeit sports jerseys, saved some money
Dad gave him $2,000
Loved to draw lines on charts when he started
Traded options using technical analysis when he started
Would make money then give it all back
Need to trade consistently to see if you're able to "escape randomness"
Lives in Florida, works for HFT firm, initially a trading assistant, lucked out getting the job
Taught himself how to program in Excel at first
Looking for edges in the data
Learning process went on for years
Need to test everything, can't just say something is "bullish" or "bearish," can't be subjective
Read "Evidence-based Technical Analysis"
Moved away from chart patterns, too tough to program and test
Technical analysis not concrete enough
To move from trading assistant to trader, he needed to know how to program
Learning to program the best trade he has ever made
Programming is a super power, find the data and test it instantly
Read textbooks, watched YouTube videos, used Coursera
C++ the language to learn
Talked to HFT firm's programmers all the time, lucky he could study on the job
First learn how to read in data and build a technical indicator from O,H,L,C data
Look at several different resources to help things click
Find people you can ask questions of when you get stuck
Aaron recommends "Stack Exchange" to find answers to questions; Dave uses it "daily"
First language you should learn is Python -- clean, simple language, lots of free libraries
Machine learning background preferred
"Fitness Function" -- run a search program for 2.5 profit factor, for example
Think of "edge" as positive expectation, "positive expectancy"
Doesn't want to give away "secret sauce"
Every rule in a strategy requires a minimum number of trades (hundreds) to test
Allow law of large numbers to play out, must flip the coin a thousand times
Split up in and out of sample data (default 35% out of sample)
Two things he likes to look at: volume and volatility
Three key points that turned his trading around, he had unrealistic expectations [other two?]
Learned about Monte Carlo simulations, recalculate the equity curve for each shuffle
Must create realistic expectations in the beginning
Where do you expect to be in the next 'n' trades? What's the distribution?
Simulate variable win rates, don't just stick with one
"Randomness happens"
Three laws that he trades by:

Risk to reward must be asymmetric, prefers trend following to mean reversion
All bets mean the same thing to your bottom line, bet sizing must be consistent for the math to play out
Make the law of large numbers work for you, edge must play out over thousands of trades

Discretionary traders who size up on certain trades are doing themselves a disservice
Trading should be boring, never exciting
Why bet five times your usual size on the eighth coin flip?
Dangerous to get an existing idea to trade more by dropping to a lower time frame, for example
Transaction costs for retail accounts make it impossible for law of large numbers to work out
Latency aribtrage has a bad rap
Trading is tough, HFT a scapegoat (but he's biased)
HFT beneficial, they do nothing predatory
Heyday of HFT over, very difficult now, strategies exposed, people jump from shop to shop
Has software he will license to the public, trading strategy search
www.buildalpha.com
Twitter: @Dburgh